原创,转发请注明出处。
MapReduce是hadoop这只大象的核心,Hadoop 中,数据处理核心就是 MapReduce 程序设计模型。一个Map/Reduce 作业(job) 通常会把输入的数据集切分为若干独立的数据块,由 map任务(task)以完全并行的方式处理它们。框架会对map的输出先进行排序, 然后把结果输入给reduce任务。通常作业的输入和输出都会被存储在文件系统中。因此,我们的编程中心主要是 mapper阶段和reducer阶段。
下面来从零开发一个MapReduce程序,并在hadoop集群上运行。
mapper代码 map.py:
import sys for line in sys.stdin: word_list = line.strip().split(' ') for word in word_list: print ' '.join([word.strip(), str(1)])
reducer代码 reduce.py:
import sys cur_word = None sum = 0 for line in sys.stdin: ss = line.strip().split(' ') if len(ss) < 2: continue word = ss[0].strip() count = ss[1].strip() if cur_word == None: cur_word = word if cur_word != word: print ' '.join([cur_word, str(sum)]) cur_word = word sum = 0 sum += int(count) print ' '.join([cur_word, str(sum)]) sum = 0
资源文件 src.txt(测试用,在集群中跑时,记得上传到hdfs上):
hello ni hao ni haoni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ao ni haoni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni haoao ni haoni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao ni hao Dad would get out his mandolin and play for the family Dad loved to play the mandolin for his family he knew we enjoyed singing I had to mature into a man and have children of my own before I realized how much he had sacrificed I had to,mature into a man and,have children of my own before.I realized how much he had sacrificed
首先本地调试查看结果是否正确,输入命令以下:
cat src.txt | python map.py | sort -k 1 | python reduce.py
命令行中输出的结果:
a 2 and 2 and,have 1 ao 1 before 1 before.I 1 children 2 Dad 2 enjoyed 1 family 2 for 2 get 1 had 4 hao 33 haoao 1 haoni 3 have 1 he 3 hello 1 his 2 how 2 I 3 into 2 knew 1 loved 1 man 2 mandolin 2 mature 1 much 2 my 2 ni 34 of 2 out 1 own 2 play 2 realized 2 sacrificed 2 singing 1 the 2 to 2 to,mature 1 we 1 would 1
通过调试发现本地调试,代码是OK的。下面扔到集群上面跑。为了方便,专门写了一个脚本 run.sh,解放劳动力嘛。
HADOOP_CMD="/home/hadoop/hadoop/bin/hadoop" STREAM_JAR_PATH="/home/hadoop/hadoop/contrib/streaming/hadoop-streaming-1.2.1.jar" INPUT_FILE_PATH="/home/input/src.txt" OUTPUT_PATH="/home/output" $HADOOP_CMD fs -rmr $OUTPUT_PATH $HADOOP_CMD jar $STREAM_JAR_PATH -input $INPUT_FILE_PATH -output $OUTPUT_PATH -mapper "python map.py" -reducer "python reduce.py" -file ./map.py -file ./reduce.py
下面解析下脚本:
HADOOP_CMD: hadoop的bin的路径 STREAM_JAR_PATH:streaming jar包的路径 INPUT_FILE_PATH:hadoop集群上的资源输入路径 OUTPUT_PATH:hadoop集群上的结果输出路径。(注意:这个目录不应该存在的,因此在脚本加了先删除这个目录。**注意****注意****注意**:若是第一次执行,没有这个目录,会报错的。可以先手动新建一个新的output目录。) $HADOOP_CMD fs -rmr $OUTPUT_PATH $HADOOP_CMD jar $STREAM_JAR_PATH -input $INPUT_FILE_PATH -output $OUTPUT_PATH -mapper "python map.py" -reducer "python reduce.py" -file ./map.py -file ./reduce.py #这里固定格式,指定输入,输出的路径;指定mapper,reducer的文件;并分发mapper,reducer角色的我们用户写的代码文件,因为集群其他的节点还没有mapper、reducer的可执行文件。
输入以下命令查看经过reduce阶段后输出的记录:
cat src.txt | python map.py | sort -k 1 | python reduce.py | wc -l
命令行中输出:43
在浏览器输入:master:50030 查看任务的详细情况。
Kind % Complete Num Tasks Pending Running Complete Killed Failed/Killed Task Attempts map 100.00% 2 0 0 2 0 0 / 0 reduce 100.00% 1 0 0 1 0 0 / 0
Map-Reduce Framework中看到这个。
Counter Map Reduce Total Reduce output records 0 0 43
证明整个过程成功。第一个hadoop程序开发结束。